You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@avalon.apache.org by Leo Simons <le...@apache.org> on 2003/01/07 12:20:12 UTC
An Avalon Roadmap?
Hi gang,
let's get everyone's agendas out on the table and figure out where we
all want to take avalon in 2003, and then lets see how much we can get
those various directions to converge. Good idea?
My idea is to kick of some discussion here, then when it gets somewhat
difficult to maintain any roadmap doc, we move the roadmap to the wiki
(keeping discussion here).
My idea is not to make something authoritive, just to figure out what
various developers want to work on, and then get any problems the PMC
might have with that (basically if it doesn't fit with the board
resolution, we either must try and get the board to accept a new
resolution or change plans) on the table too.
Starting of....some bullets to comment on (no ordering yet ;):
1. Infrastructure
-----------------
Mostly boring tasks if you ask me, but it needs doing.
- move to avalon.apache.org (think Paul's on this)
- move over mailing lists to avalon.apache.org (think this might be
difficult from an infrastructure point)
- refactor CVS (or move to svn! Apache Commons is getting svn!)
- get a better build system in place (centipede, maven or something else)
- convert all docs over to forrest (right?)
2. Procedure, policy, guideline
-------------------------------
Something I also find boring, but quite vital to the project. Hopefully
we can refer or copy-n-paste a lot from other projects.
- write avalon charter
- write avalon project docs (ie something like the jakarta 'bylaws' but
shorter and referring to "default" where possible)
- document some undocumented stuff procedure stuff like
- release plan
- voting guidelines
- committer 'internship'
3. Dealing with 'legacy' code
-----------------------------
The Avalon codebase (all cvses together) is big, too big.
- remove stale code that hasn't been released before
- move code not fitting for the avalon project under the wings of
another PMC, or perhaps to sourceforge
- deprecate ECM (but we need fortress or ubercontainer first!)
4. Work on existing code
------------------------
- do needed maintainance on "Developing with Avalon"
- do some Andy-Oliver-style example-based documentation
- maintainance releases of existing code
5. Work on new code
-------------------
This is the 2nd most important bit (I feel a charter is really needed,
too). How are we going to go about container convergence?
- Phoenix 4.3, 4.4, 4.5?
- Fortress 1, 1.1, 1.2?
- Avalon5?
- Merlin1?
- UberContainer?
- Containerkit2?
- AvalonJ2ME?
I for one don't believe J2ME wants or needs avalon, and I don't feel
like making big compromises for it. I also think it is a bad idea to put
a strong break on development of phoenix, as it is somewhat unsure where
the winds will take any other setup.
I have a need for ECM2, aka Fortress, so I'd like a release. If we don't
do a release I'll be likely running on a forked nightly.
I would rather hold on Avalon5 until we have an "ubercontainer" in place
for A4.1.
As to ubercontainer....perhaps we should just start with it and see
where it goes?
cheers,
- Leo
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Stephen McConnell <mc...@apache.org>.
Leo Simons wrote:
> Hi gang,
>
> let's get everyone's agendas out on the table and figure out where we
> all want to take avalon in 2003, and then lets see how much we can get
> those various directions to converge. Good idea?
Steve's Agenda
--------------
Context:
I have an interest in delivering next generation service and
process management solutions. Solutions I have been involved
with all based on Avalon 4. These includes a user-centric
business model of people, places and things, a minimal PKI
system (CA, RA, etc.), notification services, domain
management, adaptive security solutions (driven my some advanced
meta based systems), a solution cross-enterprise component
collaboration (also meta based), registration and discovery
services, web server and web navigation tools, and suite of
supporting systems extensions, together with the underlying
component level content here at Avalon (framework, excalibur,
sandbox).
Ambitions:
Two abmisions:
1. Understand open-source (sociology and physiology)
2. Evolution of a distributed component architecture.
Ok, the first is a little esoteric but it something I'm
interested in. The second is a little more pragmatic and
related to the work I do here in Paris.
On the pragmatic side:
* Avalon 4 is a great starting point.
* Meta/Assembly/Merlin is addressing real requirements that
I have today and is instrumental in developing thoughts
concerning forward looking strategies.
* A common container platform/architecture is right up high
on my list of priorities and a necessary ingredient.
* A common distributed container architecture is where I want
to go in not so distant future.
* Delivering what groups like Cocoon (blocks), James (Mailet),
Xxxxx (Servlet) need is really important to me because it
provides real grounding - and I do believe that we are
capable of achieving this.
Issue:
We don't have an Avalon agenda/roadmap. We are spending time
on cycles concerning product release (Cornerstone, Fortress,
Phoenix) without a clear-cut objective. I hate this. If
we were talking about releasing products in the context of a stated
Avalon agenda/roadmap then the implications would be more concrete,
measurable, and qualified and would be easier to map activities in
the greater Avalon goal.
I do understand the problem when someone has build something and
someone else is sending your off-list emails basically saying "hey
we want to use this but we can't use it until its released". I'm
sure Berin is getting this sort of thing related to Fortress, I'm
getting it with the Meta/Assembly/Merlin content. But in both
cases, the absence of a agenda/roadmap makes it difficult to
access things - it also makes it difficult for people to access
and get involved with the things I've been working on. It also
makes it difficult for me to understand how Phoenix plays into
the big picture and how I can contribute.
Roadmap:
This is most important topic for me. I really would like to see the
"Avalon Container Architecture" move ahead in the context of community
goal - I think there is a start based on the data up on the wiki. Throw
into this the really great input from users back in December and Leo's
excellent work on the context process and we are heading onto something
with substance.
I'm personally not interested in pursuing a micro Avalon (although the
factory controller scenario is definitely interesting). To be honest I
don't see where the collective bandwidth will come from to sustain this
given other subjects we are grappling with. I do think it is interesting
to have that interest present but the prior priority is the mainstream
platform.
Let's focus, deliver results, and get out press releases - show everyone
that what we can deliver seriously rocks.
For me, this means:
1. a 4.X strategy
- get Fortress released in the context of an established
agenda and roadmap - why?
* because this is a rationale ECM close
* because it can be done in the context of a roadmap
* because we need back up Berin on this
- work on a complete embedded component platform in the
3-4 month timeframe - why
* because it meets real and immediate A4 requirements
that are not addressed by anything else in Avalon
* because I figure that the amount of time it will take
to get the platform to a stable point that this both
valuable and tangible for another 1.5 - 2 years
sustainable scenario, while
* being sufficiently functional to support A5 concept
testing and validation and evolution
- forget about branding, it's not Phoenix, Merlin or
Fortress - its about a meta-model, a deployment
architecture, lifecycle strategies, lifestyle policies,
etc.
2. a 5.X strategy
- clean sheet of paper
- leveraging what we have learnt (and not so much on
what we have done)
- a context in which we can work through getting the
"Avalon rocks" platform in place
- sufficient time to do this so the end result carries
the same sentiments and protection we give to the core
framework
And qualified by the following comments:
The 4.X and 5.X strategy are concurrent. 5X needs a solid 4.X
platform backing the user community. But 4.X needs a stop point -
something close to the convergence of our current container
suite.
My feeling is that 4.X container development is about being
pragmatic and leveraging what we have as opposed to 5.X which
is a real generation jump from the 4.X mindset. 5.X should deal
with and should be measured by it's ability to meet evolving
requirements from Cocoon blocks through to James Mailet. And in the
process - I would expect to see 5.X features filtering down into
the 4.X platform and 4.X experience filtering up into the 5.X
deliverable.
At the end of the day I would anticipate 5.X seamlessly absorbing
the 4.X platform content. No preconceived ideas about dumping code
or abandoning communities - simply an elegant, advanced, adaptive
platform capable of meeting cause and fine grained component
management requirements.
Cheers, Steve.
--
Stephen J. McConnell
mailto:mcconnell@apache.org
http://www.osm.net
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
RE: An Avalon Roadmap?
Posted by Berin Loritsch <bl...@citi-us.com>.
> I don't want to discount them at all! I'm just saying that I
> myself have
> no interest in this area (for now). With "big compromise" I
> mean stuff
> like giving up xml for config. I love XML. Hope that clarifies :D
I am not proposing giving up XML for config, I am proposing that
we make a way of generating hard-coded classes for constrained
devices where parsing is problematic and only serves to slow
down the deployment.
> Hey Berin, what are you using avalon for atm, and what will
> you likely
> be using it for in the future? Are you able to talk about that?
Now? I have the project for my book, but that is it. That is
squarely a Phoenix-based project.
Future? I am maintaining a project used at U.S. consolates that
has a GUI front end, and is currently very poorly designed.
I will be refactoring it (if possible) to use Avalon and
have a good design.
The project will incorporate Web Services (either interrop with
.Net server or a J2EE server), and a component based infrastructure
in the fat client. That will help with upgrading infrastructure
as needed.
Beyond that, I can share no details....
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Leo Simons <le...@apache.org>.
Berin Loritsch wrote:
>>1. Infrastructure
>>-----------------
>>- refactor CVS (or move to svn! Apache Commons is getting svn!)
> Really? Then we might want to upgrade for Avalon!
looks like it. Greg asked the infrastructure peeps to prepare for it.
>>- write avalon project docs (ie something like the jakarta
>>'bylaws' but
>>shorter and referring to "default" where possible)
>
>
> +1 -- But not quite like 'bylaws'. Apache has a set of
> Bylaws which all projects have to abide by. It is
> more the guidelines to which we adhere (fitting within
> the bylaws).
I'm actually not sure what the term means....I ment the stuff @
http://jakarta.apache.org/site/guidelines.html
>> - committer 'internship'
>
> +1 Are you referring to the de facto standard of 6 months
> involvement in most projects before comit rights are granted?
that too. Also the 'mentoring' of new committers. For example, Peter
(the Donald variant) proposed me as a committer and after that looked
after me (cleaned up code I messed up, answered spoken and unspoken
questions...). Like a new project has a sponsoring ASF member, a new
committer can benefit from a sponsoring committer. When you think about
how a committer becomes part of the group....it's a good idea to
describe such an informal process. It makes a community transparent.
Anyway, I'll know precisely what I mean when I get to write the docs on
it :D
>>4. Work on existing code
>>------------------------
>>- do needed maintainance on "Developing with Avalon"
>
> +10
(for the people wondering or with an itch, there's conversion to forrest
left and then more doccing, like updating for new containers 'n stuff,
also Berin never got to his "part 2" of talking more about cornerstone
and phoenix; Peter explained to me some time ago how to go about the
forrest conversion and I responded mostly with shamefully not handling
it after I got angry at the tools not working)
> Merlin 2.0 (which becomes the new "Fortress")
> - Merge the best parts of Fortress into Merlin
> - Provide migration tools and compatibility.
> - Provide docs
I'll add:
- clean up merlin to a stage where it looks more like phoenix (ie
interface/impl seperation, client+bootstrap+impl jar....),
perhaps throwing in some of the existing phoenix interfaces
>>I for one don't believe J2ME wants or needs avalon, and I don't feel
>>like making big compromises for it.
>
> Our primary goal is making development easier. Simplifying.
> There are folks who would like to use J2ME, they are new to
> our fold and seem genuinely interested. I don't think we should
> discount them so quickly. We can (as I demonstrated) make
> interfaces that are J2ME friendly and do not block providing
> a compatibility layer for A4 components.
I don't want to discount them at all! I'm just saying that I myself have
no interest in this area (for now). With "big compromise" I mean stuff
like giving up xml for config. I love XML. Hope that clarifies :D
Hey Berin, what are you using avalon for atm, and what will you likely
be using it for in the future? Are you able to talk about that?
cheers,
- Leo
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Peter Donald <pe...@realityforge.org>.
On Wed, 8 Jan 2003 15:04, Berin Loritsch wrote:
> Nothing. Again, I didn't get too deep into the systems--just enough to
> post some first impressions and get some feedback from the list.
>
> Between you and Jason, I am becoming a Maven believer ;P
Maven is pretty kool. However the thing I like most about it is definetly
Jelly. It is fantastic to use. It belongs in the same category as ant and
xdoclet projects - as soon as you start using them you just can't go back! ;)
Go have a play with it - it is nice! It is still under fairly development but
once it settles I am so going to propose its adoption for all of Avalons
config/setup/doc/tool/process needs ;)
--
Cheers,
Peter Donald
"Artists can color the sky red because they know it's blue. Those of us who
aren't artists must color things the way they really are or people might
think we're stupid." -- Jules Feiffer
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Berin Loritsch <bl...@apache.org>.
Peter Donald wrote:
> On Wed, 8 Jan 2003 14:34, Berin Loritsch wrote:
>
>>>>So far Centipede and Maven are the best choices. As long as we
>>>>need to add in our own tasks like generating meta-data, or generating
>>>>bytecode, then Centipede might be the better choice for two reaons:
>>>
>>>You have got to be kidding!
>>>
>>>Maven rocks and Jelly is far more flexible way to integrate stuff. With
>>>the new features in Maven CVS there really doesn't seem to be any
>>>technical reason not to go with it.
>>
>>Have you tried centipede lately?
>
>
> A couple of weeks ago.
>
>
>>Like I said in another thread, I compared the last releases from them
>>both. If Maven has a butt-load of additional features that are not
>>in the current release, then I would like to see it in action.
>
>
> Theres some nice things been added to maven in current CVS but to be frank the
> only real one that was needed is download of plugins. Once that is in place
> the rest is nice (especially console plugin) but not critical to our usage.
>
> The flexability is in last release (which is what I am using at the moment).
> For example heres a snippet where descriptors are generated. It just runs
> after the test:compile goal. What do you think is difficult about the
> following?
>
> <postGoal name="test:compile">
> <copy todir="${maven.test.dest}" >
> <fileset dir="src/test"/>
> </copy>
> <taskdef name="metagenerate"
> classname="org.realityforge.metaclass.tools.tasks.MetaGenerateTask">
> <classpath>
> <path refid="maven.dependency.classpath"/>
> <pathelement path="${maven.build.dest}"/>
> </classpath>
> </taskdef>
>
> <metagenerate destDir="${maven.test.dest}">
> <fileset dir="${maven.test.dest}">
> <include name="**/data/*.java"/>
> </fileset>
> </metagenerate>
> </postGoal>
>
Nothing. Again, I didn't get too deep into the systems--just enough to
post some first impressions and get some feedback from the list.
Between you and Jason, I am becoming a Maven believer ;P
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Peter Donald <pe...@realityforge.org>.
On Wed, 8 Jan 2003 14:34, Berin Loritsch wrote:
> >>So far Centipede and Maven are the best choices. As long as we
> >>need to add in our own tasks like generating meta-data, or generating
> >>bytecode, then Centipede might be the better choice for two reaons:
> >
> > You have got to be kidding!
> >
> > Maven rocks and Jelly is far more flexible way to integrate stuff. With
> > the new features in Maven CVS there really doesn't seem to be any
> > technical reason not to go with it.
>
> Have you tried centipede lately?
A couple of weeks ago.
> Like I said in another thread, I compared the last releases from them
> both. If Maven has a butt-load of additional features that are not
> in the current release, then I would like to see it in action.
Theres some nice things been added to maven in current CVS but to be frank the
only real one that was needed is download of plugins. Once that is in place
the rest is nice (especially console plugin) but not critical to our usage.
The flexability is in last release (which is what I am using at the moment).
For example heres a snippet where descriptors are generated. It just runs
after the test:compile goal. What do you think is difficult about the
following?
<postGoal name="test:compile">
<copy todir="${maven.test.dest}" >
<fileset dir="src/test"/>
</copy>
<taskdef name="metagenerate"
classname="org.realityforge.metaclass.tools.tasks.MetaGenerateTask">
<classpath>
<path refid="maven.dependency.classpath"/>
<pathelement path="${maven.build.dest}"/>
</classpath>
</taskdef>
<metagenerate destDir="${maven.test.dest}">
<fileset dir="${maven.test.dest}">
<include name="**/data/*.java"/>
</fileset>
</metagenerate>
</postGoal>
--
Cheers,
Peter Donald
------------------------------------
The two secrets to success:
1- Don't tell anyone everything.
------------------------------------
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Berin Loritsch <bl...@apache.org>.
Peter Donald wrote:
> On Wed, 8 Jan 2003 02:46, Berin Loritsch wrote:
>
>>>- get a better build system in place (centipede, maven or
>>>something else)
>>
>>So far Centipede and Maven are the best choices. As long as we
>>need to add in our own tasks like generating meta-data, or generating
>>bytecode, then Centipede might be the better choice for two reaons:
>
>
> You have got to be kidding!
>
> Maven rocks and Jelly is far more flexible way to integrate stuff. With the
> new features in Maven CVS there really doesn't seem to be any technical
> reason not to go with it.
>
Have you tried centipede lately?
Like I said in another thread, I compared the last releases from them
both. If Maven has a butt-load of additional features that are not
in the current release, then I would like to see it in action.
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: An Avalon Roadmap?
Posted by Peter Donald <pe...@realityforge.org>.
On Wed, 8 Jan 2003 02:46, Berin Loritsch wrote:
> > - get a better build system in place (centipede, maven or
> > something else)
>
> So far Centipede and Maven are the best choices. As long as we
> need to add in our own tasks like generating meta-data, or generating
> bytecode, then Centipede might be the better choice for two reaons:
You have got to be kidding!
Maven rocks and Jelly is far more flexible way to integrate stuff. With the
new features in Maven CVS there really doesn't seem to be any technical
reason not to go with it.
--
Cheers,
Peter Donald
-------------------------------------------------------
To fight and conquer in all your battles is not supreme
excellence; supreme excellence consists in breaking the
enemy's resistance without fighting. - Sun Tzu, 300 B.C.
-------------------------------------------------------
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
RE: An Avalon Roadmap?
Posted by Berin Loritsch <bl...@citi-us.com>.
> From: news [mailto:news@main.gmane.org]On Behalf Of Leo Simons
>
> Hi gang,
>
> let's get everyone's agendas out on the table and figure out where we
> all want to take avalon in 2003, and then lets see how much
> we can get
> those various directions to converge. Good idea?
Great idea.
> My idea is to kick of some discussion here, then when it gets
> somewhat
> difficult to maintain any roadmap doc, we move the roadmap to
> the wiki
> (keeping discussion here).
> My idea is not to make something authoritive, just to figure out what
> various developers want to work on, and then get any problems the PMC
> might have with that (basically if it doesn't fit with the board
> resolution, we either must try and get the board to accept a new
> resolution or change plans) on the table too.
>
> Starting of....some bullets to comment on (no ordering yet ;):
+1
> 1. Infrastructure
> -----------------
> Mostly boring tasks if you ask me, but it needs doing.
>
> - move to avalon.apache.org (think Paul's on this)
> - move over mailing lists to avalon.apache.org (think this might be
> difficult from an infrastructure point)
> - refactor CVS (or move to svn! Apache Commons is getting svn!)
Really? Then we might want to upgrade for Avalon!
> - get a better build system in place (centipede, maven or
> something else)
So far Centipede and Maven are the best choices. As long as we
need to add in our own tasks like generating meta-data, or generating
bytecode, then Centipede might be the better choice for two reaons:
1) standard integration with Forrest (focus on doc generation and
site management)
2) Easy to incorporate project specific plugins (i.e. meta-gen and
family) without requiring the user to install stuff separately.
> - convert all docs over to forrest (right?)
:) I believe so. That way we can get rid of all the extra JARs
that we have simply for building docs. We can also get rid of
recursive dependencies.
> 2. Procedure, policy, guideline
> -------------------------------
> Something I also find boring, but quite vital to the project.
> Hopefully
> we can refer or copy-n-paste a lot from other projects.
>
> - write avalon charter
+1
> - write avalon project docs (ie something like the jakarta
> 'bylaws' but
> shorter and referring to "default" where possible)
+1 -- But not quite like 'bylaws'. Apache has a set of
Bylaws which all projects have to abide by. It is
more the guidelines to which we adhere (fitting within
the bylaws).
> - document some undocumented stuff procedure stuff like
> - release plan
+1
> - voting guidelines
+1 We have a general agreement on *how* to vote. The only
issue was some redefinition of PMC chair responsibilities
which we can't do. All we need to do is clean this up.
> - committer 'internship'
+1 Are you referring to the de facto standard of 6 months
involvement in most projects before comit rights are granted?
> 3. Dealing with 'legacy' code
> -----------------------------
> The Avalon codebase (all cvses together) is big, too big.
>
> - remove stale code that hasn't been released before
+1 Zip it up and remove the CVS stuff. The archive can be
placed on the server for download if it is necessary, or
moved to another project (like commons) if it is more
appropriate.
> - move code not fitting for the avalon project under the wings of
> another PMC, or perhaps to sourceforge
+1.
> - deprecate ECM (but we need fortress or ubercontainer first!)
I think Stephen will be happy with Fortress release if after
we are done with that we help him get Merlin ready. I think
that will help us understand each other better and provide
a foundation for ubercontainer cooperation. We will know
the strengths and weaknesses of each project.
> 4. Work on existing code
> ------------------------
> - do needed maintainance on "Developing with Avalon"
+10
> - do some Andy-Oliver-style example-based documentation
+1
> - maintainance releases of existing code
+1
>
> 5. Work on new code
> -------------------
> This is the 2nd most important bit (I feel a charter is
> really needed,
> too). How are we going to go about container convergence?
>
> - Phoenix 4.3, 4.4, 4.5?
> - Fortress 1, 1.1, 1.2?
> - Avalon5?
> - Merlin1?
> - UberContainer?
> - Containerkit2?
> - AvalonJ2ME?
My thoughts on this are:
Phoenix 4.x.y progression.
Fortress 1.0
- Finish docs
- Make release
Merlin 2.0 (which becomes the new "Fortress")
- Merge the best parts of Fortress into Merlin
- Provide migration tools and compatibility.
- Provide docs
Avalon5 interfaces
Ubercontainer
I think we can make Avalon5 both easier to use, and
workable on a greater variety of systems. The process
will also make Avalon more efficient and scalable which
helps on the server side and the micro client side.
Ubercontainer will be modular, possibly using AOP to
glue together crosscutting concerns.
> I for one don't believe J2ME wants or needs avalon, and I don't feel
> like making big compromises for it. I also think it is a bad
> idea to put
> a strong break on development of phoenix, as it is somewhat
> unsure where
> the winds will take any other setup.
Our primary goal is making development easier. Simplifying.
There are folks who would like to use J2ME, they are new to
our fold and seem genuinely interested. I don't think we should
discount them so quickly. We can (as I demonstrated) make
interfaces that are J2ME friendly and do not block providing
a compatibility layer for A4 components.
> I have a need for ECM2, aka Fortress, so I'd like a release.
> If we don't
> do a release I'll be likely running on a forked nightly.
I think that should be our first code priority.
> I would rather hold on Avalon5 until we have an
> "ubercontainer" in place
> for A4.1.
Honestly, I think the "ubercontainer" would be able to be made
workable with any lifecycle system we want.
> As to ubercontainer....perhaps we should just start with it and see
> where it goes?
There is that possibility too.
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>